You will own the integration of new model architectures into the inference engine and manage scaling deployments across thousands of machines. Additionally, you will build internal tooling to measure performance and optimize GPU resource utilization while maintaining high reliability.
You will design, build, and scale distributed reinforcement learning post-training systems, orchestrating complex workloads across thousands of GPUs. Additionally, you will develop reward infrastructure and evaluation tooling to ensure the stability and efficiency of large-scale RL runs.
You will design, implement, and optimize distributed training systems for large-scale multimodal models across thousands of GPUs. Additionally, you will build monitoring and debugging tools to ensure training stability, convergence, and efficient resource utilization.
You will profile and optimize GPU, CPU, and accelerator code to ensure multimodal models train efficiently and deploy at scale. This involves writing high-performance kernels, developing fused operations, and building automation tools for performance monitoring.
